Proposed Collaboration with CIRA/NPS on Smoke: Satellite and Auxiliary Data Analysis R. Husar and S. Raffuse, CAPITA Draft July 2001.

Proposed Collaboration with CIRA/NPS on Smoke:Satellite and Auxiliary Data Analysis

R. Husar and S. Raffuse, CAPITA

Draft July 2001

CIRA/NPS/CAPITA Collaboration• Smoke from prescribed and wild fires is becoming and significant issue for air quality management

due to the effects on health and visibility.

• The identification of biomass smoke as a component of the ambient aerosol is hampered by the strong variability of smoke composition and by the inadequacy of the current smoke tracers.

• Recognizing these needs and opportunities of the new analytical techniques, CIRA/NPS is embarking on significant smoke study to:

– Develop unambiguous and routine biomass smoke apportioning methodologies

– Characterize the biomass smoke physical, chemical and optical properties

• The CAPITA group is developing methods for:– Satellite-based smoke detection using multiple sensors

– Fusion of satellite observations with surface-based monitoring data

• CAPITA whishes to collaborate with CIRA/NPS in the up-coming smoke studies. This presentation: – Provides some background information relevant to the planning of these studies

– Illustrates the data, methods and some results of the smoke-related satellite and data fusion work

• The key persons involved are Rudolf Husar and Sean Raffuse. Current CAPITA funding is sufficient to get the project going. In the future, additional support will be sought.

Study Outline

• Based on conversations with W. Malm and others at CIRA, the following outline has emerged for a possible collaborative effort:

Data Collection

• Select a suitable geographic domain, preferably the Western US. • Select a time period (about 6 months) for intensive smoke study in the selected region.• For the selected period and region, perform extra filters analysis; collect auxiliary data

and conduct extensive data analysis.

Data Analysis

1. Smoke aerosol identification and apportionment (CIRA/NPS)2. Derivation of smoke physical, chemical and optical characteristics (CIRA/NPS)3. Spatio-temporal pattern of fire/smoke derived from satellite and surface data (CAPITA)4. Dynamic smoke simulation modeling and reconciliation with the observations (CIRA –

with CAPITA)

Recommendation: May-October 2000 Intensive

The suggested time domain for the intensive analysis is May-October 2000Data available

PM2.5 data from 1000 station FRM network, on hand ASOS one minute data for 250 stations is on hand for June-October 2000TERRA satellite fully operational by March 2000 (MODIS, MISER sensors)SeaWiFS satellite data available ATSR fire location data on handIMPROVE – soon?

Fire eventsIdaho-Montana Fires: August 2000California fires: August, September, 2000

Geographic Domain: Western US. However, the smoke from Idaho fires has spread over the eastern US as well, so…. This

could use some discussion.

CAPITA Effort: Spatial, Temporal and Spectral Pattern of Smoke

• Daily fire location detection (input to Monte Carlo model): – Night satellite, ATSR

– Daytime AVHRR

– Diurnal fire – GOES

• Smoke spatio-temporal pattern detection (MC model verification): – Horizontal pattern based on SeaWiFS, MODIS

– Surface concentration based on ASOS, PM2.5 FRM etc.

– Diurnal smoke variation from GOES

– Vertical pattern derived from combined surface-satellite data

• Smoke type characterization :– Classification by spectral reflectance (‘blue’, ‘yellow’, ‘white’ smoke)

– Chemical & optical, other characteristics of blue, yellow smoke (with CIRA)

Daily fire location detection with ATSR

– Large fires in Idaho and western Montana August 20 – 31, 2000

– The dots represent fires as detected by the ESA’s ATSR night satellite

– The detected fires all fall on three days.

– 8/22: red– 8/25: blue– 8/31: green

8/21 8/22 8/23

8/24 8/25 8/26 8/27

8/28 8/29 8/30 8/31

Although fires may be present, they are only detected sporadically due to cloud coverage and the satellite’s flight path

Smoke spatio-temporal pattern detection:

Horizontal pattern based on SeaWiFS, MODIS

Surface concentration based on ASOS, PM2.5 FRM etc.

Diurnal smoke variation from GOES

Vertical pattern derived from combined surface-satellite data

California Fires, Sept, 28, 1999 (SeaWiFS)

The fire locations are identified as red dots (ATSR)The smoke is yellow near the fire and bluish further away

The aerosol backscattering is retrieved from the satellite

Smoke-no fire?

Smoke & fire

Smoke & fire

California Fires, Oct, 28, 1999 (SeaWiFS)

• gwetwet

The dense smoke indicates that the fire is in N. California Sierra Mountains

However, the smoke has spread over much of the San Joaquin Basin.

SeaWiFS satellite data shows the smoke ‘hotspot’

The smoke is confined by the topography - the mountains extrude from the smoke layer.

Hence, the smoke in the basin is near the surface

MODIS fire and smoke detection – Idaho fires Aug 2000

From: Wei Min Hao, Yoram Kaufman

• MODIS data are available to the public staring March 2000.

• The MODIS data would allow daily fire/smoke analysis with high spatial and spectral resolution

GOES-Diurnal Smoke Dynamics

• The primary role of the GOES satellite is to provide the dynamic smoke pattern.

• GOES images (every 30 daytime minutes) reveal the transport as well as change in the smoke intensity over the day.

• The single visible channel makes it difficult to distinguish smoke, haze and clouds.

• The GOES data are best used in combination with daily SeaWiFS or MODIS data.

An experimental fire locations for California are provided by the U. Wisconsin, ABBA Project based on GOES 10.

Click here for the

Latest Fire locations in California/Nevada

The Real-Time fire locations are being used as source terms for a predictive smoke transport model. (Naval Res. Lab – Doug Westphal). Click here to view latest forecast loop.

GOES – Fire Detection and Modeling

AVHRR Fire and Smoke Detection – May 2001 Florida Fires

• NOAA AVHRR satellites pass over an area daily• The AVHRR sensors provide both fire location and smoke pattern

CARB Surface Aerosol Monitoring

CARB has an extensive aerosol monitoring network

TEOM (PM2.5), Nephelometer (Bscat), Coeff. Haze (COH) is monitored hourly

The FAA/NWS ASOS system has over 50 sites with 1 minute light scattering data

In principle, these aerosol data are available for on-line monitoring during smoke events

However, the real-time access and other possible problems with these data are not known.

ASOS Visibility Monitoring System• The instrumental visibility data are part of the

Automated Surface Observing System, ASOS, with 860 + sites over the country.

• The Belfort Model 6220, senses the forward light scattering (30-500) of white light, in situ. The nominal sensitivity is 17 ft to 30 miles.

• The synoptic ASOS stations distribute the data through the WMO system.

• The data are truncated into 18 distinct visual range bins: <1/4, 1/4,….,10+ miles. For smoke events with visibility well below 10 miles, this is not a severe limitation.

• The one-minute un-truncated data are stored on the on-site computer for 12 hours and in principle, it can be accessed remotely during that time.

Smoke types: blue, yellow, white

Smoke from major fires comes in different colors. The spectral characteristics (color) of different smoke types will be examined

What are the chemical, physical and optical characteristics of blue, yellow.. smoke (by CIRA)? Is the smoke color due to size, composition or optical thickness effects? Can the reflectance color be used to classify smokes?

CARB Aerosol Monitoring during a Smoke Event, Oct. 18, 1999

There are over a dozen CARB aerosol sites that are exposed to smoke in the Central Basin on Oct. 18, 1999

The 3 aerosol monitors show peak smoke concentration in the 200 g/m3 range

The ‘composition’ of the smoke varies; some puffs are high in scattering, others high in absorption (COH)

Satellite Detection of Fires over N. America

• The purpose of the next slides is to provide a climatological view of NAM fires – as detected by satellites

• This background should be useful in selecting the study region and time domain.

• At some point, the satellite fire maps will also need to be reconciled with fire maps constructed but ground-based methods of the Forest Serve.

FIRE and Norm. Diff. Veg. Index, NDVI

The ‘Northern’ zone from Alaska to Newfoundland has large fire ‘patches’, evidence of large, contiguous fires.

The ‘Northwestern’ zone (W. Canada, ID, MT, CA) is a mixture of large and small fires

The ‘Southeastern’ fire zone (TX–NC–FL) has a moderate density of uniformly distributed small fires.

The ‘Mexican’ zone over low elevation C America is the most intense fire zone, sharply separated from arid and the lush regions.

Fires are absent in arid low-vegetation areas (yellow) and over areas of heavy, moist vegetation (blue).

Fire Zones of North America

Seasonality of Fire

Dec, Jan, Feb is generally fire-free except in Mexico, and W. Canada

Mar, Apr, May is the peak fire season in Mexico and Cuba; fires occur also in Alberta-Manitoba and in OK-MO region

Jun, Jul, Aug is the peak fire season in N. Canada, Alaska and the NW US.

Sep, Oct, Nov is fire over the ‘Northwest’ and the “Southeast’

Seasonal Pattern of Fires over N. America

• The number of ATSR satellite-observed fires peaks in warm season

• Fires are random; onset and smoke amount is unpredictable

Seasonal Pattern of Fires over the ‘West’ (30N-52N, > 100W)

• The fire count gradually increases from March through May and sharply declines by June• The daily fire counts over the entire West varies by an order of magnitude during the fire season

West

Seasonal Pattern of Fires over the ‘East’ (30N-52N, < 100W)

• About 10-100 fires are recorded from March-November.• The fire counts over the East are low compared to other regions of NAM.

East

Fire Locations on Elevation Map (1996-2000)

• Over a short 5-year period, the fire locations are spatially clustered

• Fire locations over the Western US are mainly at higher elevations

Spatial Pattern and History of Fire Counts, CA, NV (1996-1999)

• The CA-NV fires occur very sporadically. Some years have virtually no fires while others have up to 1000 1 km-size ‘fire pixels’/day.

• Within a fire season, the fire events occurs sporadically, spaced in fire-free periods.

Elevation Map

Fire Locations

CaliforniaFires (1997-2000

Large fires in California are evident as clusters ‘fire pixels’

Large fires tend to occur at higher elevation mountainous areas

The fires just south of Mexicali are distributed randomly over a larger area